How does governance determine whether AI initiatives ever leave
the proof-of-concept stage?
That was the focus of our Delivery Director, Lesia Kasian, at yesterday’s Digital Telco Strategy Forum & AI & Data Analytics Meeting by ETIS.
While most organizations are already exploring AI, few have built the internal discipline to make it work at scale.
Lesia shared several lessons from telecom projects in which governance shaped the outcomes:
— AI projects fail because the processes generating that data aren’t clean or accountable.
Without a unified framework for using AI, employees turn to their own tools,
a kind of “shadow AI” that operates outside governance and validation.
As a result, AI adoption often looks active on paper, while in reality, there’s no centralized strategy or control over data quality. It creates a data-reality gap, where analytics reflect system noise rather than actual operations.
Without stable workflows, accountable teams, and validated data streams, AI can’t move beyond experiments.
— Only 5–12% of AI pilots reach production, because leadership sees risk before return.
Demonstrating measurable impact is what turns approval into sustained investment.
Once leadership understands how AI changes workflows, documentation, and control loops, the organization finally shifts from trials to execution.
— Modernization begins when governance aligns leadership, data, and delivery under one system of responsibility.
Governance connects financial validation with delivery discipline, making impact measurable and investments defensible.
Operational maturity depends on people as much as on systems: from data engineers to MLOps teams working under shared accountability.
And MLOps itself is continuous governance: retraining models, tracking metrics drift, and keeping accuracy within defined thresholds.
______
Thanks to ETIS - The Community of Telecom Professionals for the opportunity to share our perspective and learn from the community.
#DigitalTelco #AIinTelecom #ServiceModernization #A2A #MCP #ETISCommunity